Search

A multi-modal search space can be defined as having multiple attraction basins – each basin has a single local optimum which is reached from all points in that basin when greedy local search is used. Optimization in multi-modal search spaces can then be viewed as a two-phase process. The first phase is exploration in which the most promising attraction basin is identified. The second phase is exploitation in which the best solution (i.e. the local optimum) within the previously identified attraction basin is attained. The goal of thresheld convergence is to improve the performance of search techniques during the first phase of exploration. The effectiveness of thresheld convergence has been demonstrated through applications to existing metaheuristics such as particle swarm optimization and differential evolution, and through the development of novel metaheuristics such as minimum population search and leaders and followers.

Computer modeling of protein ligand interactions is one of the most important phases in a drug design process. Part of the process involves the optimization of highly multi-modal objective (scoring) functions. This research presents the Minimum Population Search heuristic as an alternative for solving these global unconstrained optimization problems. To determine the effectiveness of Minimum Population Search, a comparison with seven state-of-the-art search heuristics is performed. Being specifically designed for the optimization of large scale multi modal problems, Minimum Population Search achieves excellent results on all of the tested complexes, especially when the amount of available function evaluations is strongly reduced. A first step is also made toward the design of hybrid algorithms based on the exploratory power of Minimum Population Search. Computational results show that hybridization leads to a further improvement in performance.

El cómputo de alto rendimiento es una necesidad para el desarrollo de investigaciones con grandes volúmenes de datos. La creciente demanda de este tipo de resultados ha impulsado a varios centros de investigación a poner en funcionamiento recursos de cómputo de alto rendimiento. En Cuba no existe una solución definitiva que permita a todos los centros de investigación disponer de los recursos de cómputo necesarios para desarrollar sus proyectos. Este trabajo propone el empleo de un clúster de computadoras de la Universidad de Griffith a través de una interfaz basada en el correo electrónico. Esta solución permite disponer de recursos de cómputo de alto rendimiento sin necesidad de una alta conectividad. Como caso de estudio se analizan los resultados obtenidos en un proyecto de optimización global en grandes dimensiones desarrollado en la Universidad de La Habana. Para experimentos con un mes de duración (en una computadora estándar) los resultados muestran que al utilizar el recurso de alto rendimiento es posible alcanzar un incremento en el rendimiento relativo superior al 1300%.

A multi-modal search space can be defined as having multiple attraction basins – each basin has a single local optimum which is reached from all points in that basin when greedy local search is used. Optimization in multi-modal search spaces can then be viewed as a two-phase process. The first phase is exploration in which the most promising attraction basin is identified. The second phase is exploitation in which the best solution (i.e. the local optimum) within the previously identified attraction basin is attained. The goal of thresheld convergence is to improve the performance of search techniques during the first phase of exploration. The effectiveness of thresheld convergence has been demonstrated through applications to existing metaheuristics such as particle swarm optimization and differential evolution, and through the development of novel metaheuristics such as minimum population search and leaders and followers.

Computer modeling of protein-ligand interactions is one of the most important phases in a drug design process. Part of the process involves the optimization of highly multi-modal objective (scoring) functions. This research presents the Minimum Population Search heuristic as an alternative for solving these global unconstrained optimization problems. To determine the effectiveness of Minimum Population Search, a comparison with seven state-of-the-art search heuristics is performed. Being specifically designed for the optimization of large scale multi-modal problems, Minimum Population Search achieves excellent results on all of the tested complexes, especially when the amount of available function evaluations is strongly reduced. A first step is also made toward the design of hybrid algorithms based on the exploratory power of Minimum Population Search. Computational results show that hybridization leads to a further improvement in performance.

Resumen El cómputo de alto rendimiento es una necesidad para el desarrollo de investigaciones con grandes volúmenes de datos. La creciente demanda de este tipo de resultados ha impulsado a varios centros de investigación a poner en funcionamiento recursos de cómputo de alto rendimiento. En Cuba no existe una solución definitiva que permita a todos los centros de investigación disponer de los recursos de cómputo necesarios para desarrollar sus proyectos. Este trabajo propone el empleo de un clúster de computadoras de la Universidad de Griffith a través de una interfaz basada en el correo electrónico. Esta solución permite disponer de recursos de cómputo de alto rendimiento sin necesidad de una alta conectividad. Como caso de estudio se analizan los resultados obtenidos en un proyecto de optimización global en grandes dimensiones desarrollado en la Universidad de La Habana. Para experimentos con un mes de duración (en una computadora estándar) los resultados muestran que al utilizar el recurso de alto rendimiento es posible alcanzar un incremento en el rendimiento relativo superior al 1300%. Abstract: Research with large volumes of data usually require access to High Performance Computing. The increasing demand for this kind of research has led many institutions to develop their own computer clusters. However, in Cuba there is no definitive solution for the High Performance Computing requirements of institutions such as the University of Havana. The expenses of building a computer cluster disallows many institutions to have their own, while the low connectivity limits the use of international high performance computing services. This research presents an alternative solution based on the development of an email-based platform for accessing a computer cluster at Griffith University in Australia. This new communication interface has been successfully used on a Large Scale Optimization research project at the University of Havana. Computational results show that by using the remote cluster it is possible to improve upon the performance of a personal computer by up to 1300%.